globalsea–airco fluxbasedonclimatologicalsurfaceocean ...csweeney/papers/taka2002.pdf ·...

22
Deep-Sea Research II 49 (2002) 1601–1622 Global sea–air CO 2 flux based on climatological surface ocean pCO 2 , and seasonal biological and temperature effects Taro Takahashi a, *, Stewart C. Sutherland a , Colm Sweeney a , Alain Poisson b , Nicolas Metzl b , Bronte Tilbrook c , Nicolas Bates d , Rik Wanninkhof e , Richard A. Feely f , Christopher Sabine f , Jon Olafsson g , Yukihiro Nojiri h a Lamont-Doherty Earth Observatory of Columbia University, Palisades, NY 10964, USA b Laboratoire de Biog ! eochimie et Chimie Marines, Universite Pierre et Marie Curie, Case 134 4 place Jussieu, 75252 Paris Cedex 05, France c Antarctic CRC and CSIRO Marine Research, Hobart, Tasmania 7001, Australia d Bermuda Biological Station for Research, Ferry Reach GE 01, Bermuda e AOML, National Oceanographic and Atmospheric Administration, 4301 Rickenbacker Causeway, Miami, FL 33149, USA f PMEL, National Oceanographic and Atmospheric Administration, 7600 Sand Point Way, NE, Seattle, WA 98115, USA g University of Iceland and Marine Research Institute, Skulagata 4, 121, Reykjavik, Iceland h Global Warming Mechanism Laboratory, National Institute for Environmental Studies, Tsukuba, Ibaraki 305-0053, Japan Abstract Based on about 940,000 measurements of surface-water pCO 2 obtained since the International Geophysical Year of 1956–59, the climatological, monthly distribution of pCO 2 in the global surface waters representing mean non-El Ni * no conditions has been obtained with a spatial resolution of 41 51 for a reference year 1995. The monthly and annual net sea–air CO 2 flux has been computed using the NCEP/NCAR 41-year mean monthly wind speeds. An annual net uptake flux of CO 2 by the global oceans has been estimated to be 2.2 (+22% or 19%) Pg C yr 1 using the (wind speed) 2 dependence of the CO 2 gas transfer velocity of Wanninkhof (J. Geophys. Res. 97 (1992) 7373). The errors associated with the wind-speed variation have been estimated using one standard deviation (about72ms 1 ) from the mean monthly wind speed observed over each 41 51 pixel area of the global oceans. The new global uptake flux obtained with the Wanninkhof (wind speed) 2 dependence is compared with those obtained previously using a smaller number of measurements, about 250,000 and 550,000, respectively, and are found to be consistent within70.2 Pg C yr 1 . This estimate for the global ocean uptake flux is consistent with the values of 2.070.6 Pg C yr 1 estimated on the basis of the observed changes in the atmospheric CO 2 and oxygen concentrations during the 1990s (Nature 381 (1996) 218; Science 287 (2000) 2467). However, if the (wind speed) 3 dependence of Wanninkhof and McGillis (Res. Lett. 26 (1999) 1889) is used instead, the annual ocean uptake as well as the sensitivity to wind-speed variability is increased by about 70%. A zone between 401 and 601 latitudes in both the northern and southern hemispheres is found to be a major sink for atmospheric CO 2 . In these areas, poleward-flowing warm waters meet and mix with the cold subpolar waters rich in nutrients. The pCO 2 in the surface water is decreased by the cooling effect on warm waters and by the biological drawdown of pCO 2 in subpolar waters. High wind speeds over these low pCO 2 waters increase the CO 2 uptake rate by the ocean waters. *Corresponding author. Fax: +1-845-365-2312. E-mail address: [email protected] (T. Takahashi). 0967-0645/02/$ - see front matter r 2002 Published by Elsevier Science Ltd. PII:S0967-0645(02)00003-6

Upload: buithuy

Post on 10-Feb-2019

214 views

Category:

Documents


0 download

TRANSCRIPT

Deep-Sea Research II 49 (2002) 1601–1622

Global sea–air CO2 flux based on climatological surface oceanpCO2, and seasonal biological and temperature effects

Taro Takahashia,*, Stewart C. Sutherlanda, Colm Sweeneya, Alain Poissonb,Nicolas Metzlb, Bronte Tilbrookc, Nicolas Batesd, Rik Wanninkhofe,

Richard A. Feelyf, Christopher Sabinef, Jon Olafssong, Yukihiro Nojirih

aLamont-Doherty Earth Observatory of Columbia University, Palisades, NY 10964, USAbLaboratoire de Biog!eochimie et Chimie Marines, Universite Pierre et Marie Curie, Case 134 4 place Jussieu,

75252 Paris Cedex 05, FrancecAntarctic CRC and CSIRO Marine Research, Hobart, Tasmania 7001, Australia

dBermuda Biological Station for Research, Ferry Reach GE 01, BermudaeAOML, National Oceanographic and Atmospheric Administration, 4301 Rickenbacker Causeway, Miami, FL 33149, USAfPMEL, National Oceanographic and Atmospheric Administration, 7600 Sand Point Way, NE, Seattle, WA 98115, USA

gUniversity of Iceland and Marine Research Institute, Skulagata 4, 121, Reykjavik, IcelandhGlobal Warming Mechanism Laboratory, National Institute for Environmental Studies, Tsukuba, Ibaraki 305-0053, Japan

Abstract

Based on about 940,000 measurements of surface-water pCO2 obtained since the International Geophysical Year of

1956–59, the climatological, monthly distribution of pCO2 in the global surface waters representing mean non-El Ni *no

conditions has been obtained with a spatial resolution of 41� 51 for a reference year 1995. The monthly and annual net

sea–air CO2 flux has been computed using the NCEP/NCAR 41-year mean monthly wind speeds. An annual net uptake

flux of CO2 by the global oceans has been estimated to be 2.2 (+22% or �19%)Pg Cyr�1 using the (wind speed)2

dependence of the CO2 gas transfer velocity of Wanninkhof (J. Geophys. Res. 97 (1992) 7373). The errors associated

with the wind-speed variation have been estimated using one standard deviation (about72m s�1) from the mean

monthly wind speed observed over each 41� 51 pixel area of the global oceans. The new global uptake flux obtained

with the Wanninkhof (wind speed)2 dependence is compared with those obtained previously using a smaller number of

measurements, about 250,000 and 550,000, respectively, and are found to be consistent within70.2 Pg Cyr�1. This

estimate for the global ocean uptake flux is consistent with the values of 2.070.6 Pg Cyr�1 estimated on the basis of the

observed changes in the atmospheric CO2 and oxygen concentrations during the 1990s (Nature 381 (1996) 218; Science

287 (2000) 2467). However, if the (wind speed)3 dependence of Wanninkhof and McGillis (Res. Lett. 26 (1999) 1889) is

used instead, the annual ocean uptake as well as the sensitivity to wind-speed variability is increased by about 70%.

A zone between 401 and 601 latitudes in both the northern and southern hemispheres is found to be a major sink for

atmospheric CO2. In these areas, poleward-flowing warm waters meet and mix with the cold subpolar waters rich in

nutrients. The pCO2 in the surface water is decreased by the cooling effect on warm waters and by the biological

drawdown of pCO2 in subpolar waters. High wind speeds over these low pCO2 waters increase the CO2 uptake rate by

the ocean waters.

*Corresponding author. Fax: +1-845-365-2312.

E-mail address: [email protected] (T. Takahashi).

0967-0645/02/$ - see front matter r 2002 Published by Elsevier Science Ltd.

PII: S 0 9 6 7 - 0 6 4 5 ( 0 2 ) 0 0 0 0 3 - 6

The pCO2 in surface waters of the global oceans varies seasonally over a wide range of about 60% above and below

the current atmospheric pCO2 level of about 360matm. A global map showing the seasonal amplitude of surface-water

pCO2 is presented. The effect of biological utilization of CO2 is differentiated from that of seasonal temperature changes

using seasonal temperature data. The seasonal amplitude of surface-water pCO2 in high-latitude waters located

poleward of about 401 latitude and in the equatorial zone is dominated by the biology effect, whereas that in the

temperate gyre regions is dominated by the temperature effect. These effects are about 6 months out of phase.

Accordingly, along the boundaries between these two regimes, they tend to cancel each other, forming a zone of small

pCO2 amplitude. In the oligotrophic waters of the northern and southern temperate gyres, the biology effect is about

35matm on average. This is consistent with the biological export flux estimated by Laws et al. (Glob. Biogeochem.

Cycles 14 (2000) 1231). Small areas such as the northwestern Arabian Sea and the eastern equatorial Pacific, where

seasonal upwelling occurs, exhibit intense seasonal changes in pCO2 due to the biological drawdown of CO2. r 2002

Published by Elsevier Science Ltd.

Resume

Une distribution climatologique mensuelle moyenne de pCO2 de surface de l’oc!ean mondial a !et!e !etablie pour l’ann!ee

1995 (pour les ann!ees sans El Ni *no) "a partir de 940.000 mesures dans les eaux de surface, acquises depuis l’Ann!ee

G!eophysique Internationale (1956–59), avec une r!esolution spatiale de 41� 51. Les flux de CO2 air-mer ont !et!e ensuite

calcul!es, "a l’!echelle mensuelle et annuelle, en utilisant la distribution des vents mensuels du NCEP/NCAR moyenn!es sur

41 ans. Le flux net oc!eanique global de CO2 a !et!e estim!e "a un puits de 2.2 (+22% ou –19%) Pg Can�1 en utilisant

l’!equation de Wanninkhof (Geophys. Res. 97 (1992) 7373) dans laquelle le coefficient de transfert (ou la vitesse de

piston) du CO2 est exprim!e en fonction du carr!e de la vitesse du vent; les incertitudes associ!ees "a la vitesse du vent ont!et!e calcul!ees en utilisant une erreur standard de 72m s�1 pour un vent moyen observ!e sur chaque pixel de 41� 51 de

l’oc!ean global. Le nouveau flux global ainsi calcul!e (carr!e de la vitesse du vent dans le coefficient de transfert) est "a

0.2 Pg Can�1 pr"es, en accord avec nos estimations ant!erieures bas!ees respectivement sur 250.000 et 550.000 mesures

seulement. Ce r!esultat est !egalement compatible avec les estimations de 2.070.6 PgC an�1 de Keeling et al. (Nature 381

(1996) 218) et de Battle et al. (Science 287 (2000) 2467) bas!ees sur la variation des teneurs en CO2 et oxyg"ene

atmosph!eriques pendant la d!ecennie des ann!ees 90. Cependant l’utilisation de l’!equation de Wanninkhof et McGillis

(Res. Lett. 26 (1999) 1889), faisant intervenir le coefficient de transfert de CO2 exprim!e en fonction du cube de la vitesse

du vent, donne des valeurs du flux annuel global ainsi que la sensibilit!e "a la variation de la vitesse du vent plus !elev!ees de

70%.

Le puits majeur de CO2 atmosph!erique est situ!e dans la bande de latitude comprise entre 40 et 601 pour les deux

h!emisph"eres, o "u les eaux chaudes se dirigeant vers vers le p #ole convergent et se m!elangent avec des eaux froides

subpolaires riches en sels nutritifs. La diminution de pCO2 des eaux de surface subpolaires est due "a la fois "a leur

refroidissement et "a la pompe biologique. L’intensit!e des vents dans ces r!egions o "u pCO2 est faible en surface intensifie

encore le pompage de CO2.

La variation saisonni"ere de pCO2 dans les eaux de surface de l’oc!ean global peut atteindre760% de la valeur actuelle

de pCO2 atmosph!erique moyen (360 matm). Une carte de l’amplitude de la variation saisonni"ere de pCO2 dans les eaux

de surface ainsi que l’effet de l’utilisation du CO2 par la production biologique et l’influence de la variation saisonni"ere

de la temp!erature (estim!ee en utilisant des donn!ees saisonni"eres de temp!erature) sont pr!esent!ees s!epar!ement. La

variation saisonni"ere de pCO2 des eaux de surface situ!ees entre les p #oles et environ 401 et celles des r!egions !equatoriales

est principalement due "a l’effet de la pompe biologique, alors que dans les tourbillons des r!egions centrales c’est l’effet

thermique qui domine, ces deux effets !etant d!ephas!es d’environ 6 mois. En cons!equence, dans les r!egions fronti"eres de

ces deux zones les effets thermique et biologique se compensent, ce qui engendre des variations saisonni"eres de pCO2 de

faible amplitude. Dans les eaux oligotrophes des tourbillons centraux des r!egions temp!er!ees, l’influence de la biologie

est estim!e "a 35matm en moyenne, ce qui est compatible avec le flux biologique export!e estim!e par Laws et al. (Glob.

Biogeochem. Cycles 14 (2000) 1231). L’importance de l’effet biologique sur pCO2 d "u "a l’intensit!e des remont!ees

saisonni"eres d’eaux a !et!e mis en !evidence dans des r!egions de surface limit!ees du nord-ouest de la mer d’Arabie et du

Pacifique est-!equatorial.

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–16221602

1. Introduction

The difference between the pCO2 in the surfaceocean water and that in the overlying airrepresents the thermodynamic driving potentialfor the CO2 gas transfer across the sea surface. Thedirection of the net transfer of CO2 is governed bythe pCO2 differences, and the magnitude of the netsea–air CO2 flux may be expressed as a product ofthe pCO2 difference and the sea–air CO2 gastransfer velocity, which may be parameterized as afunction of wind speed. The flux values thusestimated depend on the wind field used and theformulation for the wind-speed dependence on thegas transfer velocity. In this paper, we will presenta set of the climatological monthly distributionfield of surface-water pCO2 over the global oceans,which has been improved using an expanded dataset consisting of about 940,000 pCO2 measure-ments. The effects of wind fields and the wind-speed dependence for the gas transfer velocity onthe estimated CO2 flux will be discussed using theNCEP/NCAR wind-speed data averaged over 41years and the single year data for 1995.Over the global oceans, the seasonal and

geographical variation of surface-water pCO2 ismuch greater than that of atmospheric pCO2, andhence the direction and magnitude of the sea–airCO2 transfer flux are mainly regulated by theoceanic pCO2. The pCO2 in surface-ocean water isknown to vary geographically and seasonally overa wide range between about 150 and 550 matm,about 60% below and above the mean atmo-spheric pCO2 of about 360 matm (or about370 ppm mole fraction in dry air) in the year2000. The pCO2 in mixed-layer water that ex-changes CO2 directly with the atmosphere isaffected by seasonal changes in temperature, totalCO2 concentration, and alkalinity. While the watertemperature is primarily regulated by physicalprocesses (i.e. solar energy input, sea–air heatexchanges, and mixed-layer thickness), the lattertwo are primarily controlled by biological pro-cesses (i.e. photosynthesis, respiration and calcifi-cation) and by upwelling of subsurface watersenriched in respired CO2 and nutrients. We presenta method for distinguishing the seasonal biologicaleffect on surface-water pCO2 from the effect of

seasonal temperature changes, and discuss therelative significance of these effects in differentregions of the global oceans.

2. Climatological distribution of surface-water

pCO2

2.1. Database

The climatological distribution of the surface-water pCO2 for each month over the global oceanshas been computed using about 940,000 measure-ments for pCO2 in surface waters made since theInternational Geophysical Year, 1956–59. The sizeof our present database has been increased fromabout 250,000 used in Takahashi et al. (1997) toabout 550,000 used in Takahashi et al. (1999) andthen to the present size. The data source prior to1996 is described in Takahashi et al. (1997). Thenewer data sets used in this study include Sabineet al. (1997), Bates et al. (1998), Bates (2001), Feelyet al. (1999), Metzl et al. (1995, 1999), Murphyet al. (1998), Nojiri, (2001a, b), Rubin et al. (1998),Rubin (2000), Sabine and Key (1998), Sabine et al.(1997, 1999, 2000), Sweeney (2000), Sweeney et al.(2000a, b), Takahashi and Goddard (1998), Taka-hashi et al. (2000), and the data files assembled atthe laboratories of the authors of this paper.

2.2. Computational method

The space–time interpolation method used forthis paper is based on the scheme developed byTakahashi et al. (1993). However, since a numberof improvements have been made, the methods ofdata selection and changes in the computationalprocedures will be explained briefly below.First, since pCO2 in the equatorial Pacific

surface waters changes drastically during El Ni *noperiods, the observations made in the equatorialPacific between 101N and 101S during El Ni *noevents have been excluded from the data set. Thus,the results shown in this paper represent theclimatological distributions under non-El Ni *noconditions.Second, since the surface-water pCO2 has been

changing with time in response to the atmospheric

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–1622 1603

pCO2 increase, the measurements made in differ-ent years need to be corrected to a single referenceyear (arbitrarily chosen to be 1995) on the basis ofthe following two lines of evidence:(a) Surface waters in subtropical gyres mix

vertically at slow rates with subsurface watersdue to the presence of strong stratification at thebase of the winter-mixed layer. This results in along contact time with the atmosphere to exchangeCO2. Therefore, the pCO2 in these warm watersfollows the increasing trend of atmospheric CO2 asdemonstrated by Inoue et al. (1995), Feely et al.(1999), and Bates (2001). Accordingly, the pCO2

measured in a given month and year is corrected tothe same month of the reference year 1995 usingchanges in the atmospheric CO2 concentrationoccurred during this period. Oceanic pCO2 mea-surements made after the beginning of 1979 havebeen corrected to 1995 using the atmospheric CO2

concentration data from the GLOBALVIEW-CO2 database (2000), in which the zonal meanatmospheric concentrations (for each 0.05 in sineof latitude) within the planetary boundary layerare summarized for each month since 1979–2000.Pre-1979 oceanic pCO2 data were corrected to1979 using the annual mean trend for the globalmean atmospheric CO2 concentration constructedfrom the Mauna Loa data of Keeling and Whorf(2000), and then from 1979 to 1995 using theGLOBALVIEW-CO2 database. Measurementsfor pCO2 made in the following areas have beencorrected for the time of observation; 451N–501Sin the Atlantic Ocean, north of 501S in the IndianOcean, 401N–501S in the western Pacific west ofthe date line, and 401N–601S in the eastern Pacificeast of the date line.(b) In contrast to subtropical gyres, surface

waters in high-latitude regions are mixed convec-tively with deep waters during the fall–winterseasons, and their CO2 properties tend to remainunchanged from year to year, reflecting those ofdeep waters in which the effect of increasedatmospheric CO2 is diluted to undetectable levelsas observed by Takahashi et al. (1997). Hence, nocorrection is necessary for the year of measure-ments.These pCO2 values adjusted to the reference

year 1995 are binned into each of 750,000 pixels

[=(72 pixels along the E–W direction)� (41 alongthe N–S direction)� (365 days)], which representthe global 41� 51 grid for each day in a singlevirtual calendar year. Since pCO2 measurementshave been made only sparsely in each year, it isnecessary to pool the data collected over 40 yearsinto a single reference year in order to have asufficient spatial and seasonal coverage over theglobal oceans.On the basis of the database thus assembled,

mean monthly global distributions of pCO2 havebeen constructed using an interpolation methodbased on a lateral 2-dimensional advection–diffu-sion transport equation for surface-ocean water(Takahashi et al., 1995, 1997). The monthly fieldfor lateral advection of global surface ocean watersobtained by Toggweiler et al. (1989) is used and,for the lateral diffusive transport, a constant valueof 2000m2 s�1 (Thiele et al., 1986; Jenkins, 1991) isused. The equation is solved iteratively using afinite-difference algorithm with a time step of 1day through the reference year 1995. The compu-tational domains are joined at December 31, 1995,with January 1, 1995, and along the primemeridian to ensure continuity in time and space.Typically, several thousand iterations are neces-sary before solutions are converged. The solutionsthus obtained give surface-water pCO2 values for41� 51 pixels where no observations exist, whilethe observed values are explicitly satisfied.Although the solutions give daily distributions,monthly mean distributions for the surface-waterpCO2 have been computed and are presented inFig. 1. It should be pointed out that in the earlierpapers (Takahashi et al., 1997, 1999), the sea–airpCO2 difference, DpCO2, was interpolated, insteadof seawater pCO2, using the transport equation forsurface ocean waters. This assumes implicitly thatatmospheric CO2 in the overlying air was trans-ported in the same manner as surface-oceanwaters. However, this does not represent theactual transport of the atmosphere correctly.Hence, in the present work, we applied theocean-transport equation for interpolating sur-face-ocean pCO2 data.In order to test the validity of our interpolation

scheme, we computed monthly temperature fieldsfor the world oceans using water temperature

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–16221604

values observed concurrently with pCO2 in surfacewaters, and compared them with the climatologi-cal mean monthly SST fields given in the World

Ocean Database (1998). For the observed 8938monthly pixels, the mean difference between ourpixel values and the World Ocean Database values

Climatological pCO2 in Surface Water [940K] for February 1995

30˚

30˚

60˚

60˚

90˚

90˚

120˚

120˚

150˚

150˚

180˚

180˚

150˚

150˚

120˚

120˚

90˚

90˚

60˚

60˚

30˚

30˚

80˚ 80˚70˚ 70˚

60˚ 60˚50˚ 50˚

40˚ 40˚

30˚ 30˚

20˚ 20˚

10˚ 10˚

0˚ 0˚

10˚ 10˚

20˚ 20˚

30˚ 30˚

40˚ 40˚50˚ 50˚

60˚ 60˚70˚ 70˚

80˚ 80˚

Climatological pCO2 in Surface Water [940K] for August 1995

30˚

30˚

60˚

60˚

90˚

90˚

120˚

120˚

150˚

150˚

180˚

180˚

150˚

150˚

120˚

120˚

90˚

90˚

60˚

60˚

30˚

30˚

80˚ 80˚70˚ 70˚

60˚ 60˚50˚ 50˚

40˚ 40˚

30˚ 30˚

20˚ 20˚

10˚ 10˚

0˚ 0˚

10˚ 10˚

20˚ 20˚

30˚ 30˚

40˚ 40˚50˚ 50˚

60˚ 60˚70˚ 70˚

80˚ 80˚

100 200 225 250 275 300 325 350 375 400 425 450 475 650

Seawater pCO2 (µatm) at SST

Fig. 1. Climatological mean distribution of surface-water pCO2 in February (top) and August (bottom) in the reference year 1995.

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–1622 1605

is �0.0571.401C; and for the observed andinterpolated 21,108 monthly pixels, the meandifference is 0.2871.641C. The mean differencefor the interpolated global SST field includes zerowell within one standard deviation. The meandifference of –0.051C for the observed pixelssuggests that our pCO2 observations representdistributions consistent with the climatologicalSST values. However, if 0.281C is taken as ameasure for a systematic error introduced byimperfections of our interpolation scheme, itshould correspond to an error of 73 matm inseawater pCO2, on the basis that the seasonaltemperature effect on ocean-surface-water pCO2 isin a range between +4 and �4% 1C�1 over theglobal oceans (Takahashi et al., 1993). While thetemperature effect on pCO2 in isochemical condi-tions (qln pCO2/qT) is +4.23% 1C�1 (Takahashiet al., 1993), the temperature dependence changesover the range indicated above because seasonaltemperature changes in mixed-layer water areaccompanied with an increase (by deep-waterupwelling and sea–air CO2 exchange) or decrease(by photosynthesis and sea–air CO2 exchange) inthe total CO2 concentration.

2.3. Distribution of surface-water pCO2

The mean monthly pCO2 values in Augustand February for the reference year 1995 areshown in Fig. 1 as examples of the 12 monthlydistributions obtained. These monthly distribu-tions represent climatological mean distributionsfor non-El Ni *no conditions based on measure-ments of surface-water pCO2 made from 1958through 2000.Fig. 1 shows that low pCO2 areas (blue and

purple) are found in the high-latitude areas in theSouthern Ocean during the austral summer(February, the upper panel) and in the subarcticand arctic Atlantic Ocean during the northernsummer (August, the lower panel). These areasrepresent an intense sink for atmosphericCO2, which is attributed primarily to thephotosynthetic utilization of CO2. The broad lowpCO2 areas (dark and light blue) which are foundin the mid-latitude areas of the North Pacificand North Atlantic during the northern winter

(February, the upper panel), are due primarilyto the cooling of the warm Kuroshio andGulf Stream waters, respectively, as they flow tohigher latitudes. Similar features are found inthe mid-latitude areas of the southernhemisphere oceans during the austral winter(August, the lower panel). The high pCO2

values (yellow and red) observed in the easternequatorial Pacific are a product of localupwelling of CO2-rich waters along the equatorand advection of CO2-enriched waters from theSouth American coast (Feely et al., 1999;Etcheto et al., 1999). The pCO2 value in theequatorial zone of the Pacific decreases fromeast to west as a result of biological utilizationof CO2, reduced upwelling, and loss of CO2 toair by air–sea gas exchange (Quay, 1987).High pCO2 values are also observed in thesubarctic northwestern Pacific during the northernwinter (February, the upper panel) and in theArabian Sea during the northern summer (thelower panel). The former is attributed to theconvective mixing of subsurface waters in thesubarctic Pacific during winter, and the latter tothe monsoon-induced upwelling in the ArabianSea during summer.

3. Net sea–air CO2 flux

3.1. Computational method

Net sea–air CO2 flux, F ; can be estimatedusing: F ¼ k � a � ðDpCO2Þsea2air; where k is theCO2 gas transfer velocity, a is the solubility of CO2

in seawater (Weiss, 1974), and (DpCO2)sea–air is thesea–air pCO2 difference. The sea–air pCO2 differ-ence is computed using the mean monthly pCO2

values in surface waters obtained in this study andthe atmospheric pCO2 computed using the zonalmean CO2 concentrations in the dry atmospherefor 1995 reported by the GLOBALVIEW-CO2(2000). The climatological mean barometric pres-sure (Pb) (Atlas of Surface Marine Data, 1994)and equilibrium water vapor pressure (Pw) atclimatological surface water temperature andsalinity (World Ocean Database, 1998) are usedfor computing the atmospheric pCO2 in the

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–16221606

relationship

ðpCO2Þair¼ ðCO2 conc:ÞairðPb� PwÞ:

3.2. Global distribution of the net sea–air CO2 flux

Table 1 shows the results of net sea–air CO2 fluxcomputed for regional and global oceans using the

CO2 gas transfer velocity vs. wind-speed relationformulated by Wanninkhof (1992) and the NCEP-41-year mean monthly wind speed over each pixelarea. The results of earlier studies (Takahashi et al.,1997, 1999) have been corrected using the samewind speeds and the Wanninkhof (1992) relation-ship. Although the number of pCO2 measurementsin the database has increased from about 250,000

Table 1

The net sea–air CO2 flux computed using the climatological sea–air pCO2 difference, and the NCEP/NCAR 41-year mean monthly

wind speed and the wind speed dependence of CO2 gas transfer velocity by Wanninkhof (1992; named W92 in the table)

Lat. band DpCO2 data Gas trans./

wind data

Pacific Ocean

(Pg Cyr�1)

Atlantic

Ocean

(Pg Cyr�1)

Indian Ocean

(Pg Cyr�1)

Southern

Ocean

(Pg Cyr�1)

Global

Ocean

(Pg Cyr�1)

N of 501N This work W92/41-yr +0.01 �0.40 F F �0.39T99 W92/41-yr �0.02 �0.48 F F �0.49T97 W92/41-yr +0.01 �0.48 F F �0.47

141N–501N This work W92/41-yr �0.64 �0.34 +0.07 F �0.92T99 W92/41-yr �0.62 �0.32 +0.06 F �0.87T97 W92/41-yr �0.55 �0.37 +0.05 F �0.87

141N–141S This work W92/41-yr +0.74 +0.15 +0.18 F +1.07

T99 W92/41-yr +0.73 +0.18 +0.15 F +1.06

T97 W92/41-yr +0.78 +0.15 +0.18 F +1.11

141S–501S This work W92/41-yr �0.51 �0.33 �0.67 F �1.51T99 W92/41-yr �0.48 �0.27 �0.79 F �1.54T97 W92/41-yr �0.46 �0.29 �0.57 F �1.32

S of 501S This work W92/41-yr F F F �0.47 �0.47T99 W92/41-yr F F F �0.59 �0.59T97 W92/41-yr F F F �0.42 �0.42F F F F F F F

Oceanic This work W92/41-yr �0.40 �0.92 �0.43 �0.47 �2.22Regions T99 W92/41-yr �0.39 �0.88 �0.58 �0.59 �2.44

T97 W92/41-yr �0.22 �0.99 �0.34 �0.42 �1.97

Regional This work W92/41-yr 18 41 19 21 100

Flux (%) T99 W92/41-yr 16 36 24 24 100

T97 W92/41-yr 12 50 17 21 100

Areas (106 km2) F 151.6 72.7 53.2 31.7 309.1

(%) F 49.0 23.5 17.2 10.2 100

The fluxes reported by Takahashi et al. (1999; named T99) and Takahashi et al. (1997; named T97) have been corrected using the

NCEP/NCAR 41-year mean monthly wind speeds over each 41� 51 pixel area (named 41-yr). The sea–air pCO2 difference, DpCO2,

used in this work is based on about 940,000 surface water pCO2 measurements corrected to the reference year 1995; that in T99 is based

on about 550,000 surface water pCO2 measurements corrected to the reference year 1995; and that in T97 is based on about 250,000

measurements corrected to 1990. See text for the interpolation method used in T97 and T99

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–1622 1607

for the 1997 study to about 550,000 for the 1999study and then to 940,000 for this study, theregional net sea–air pCO2 flux values are found tobe consistent within 710% on the average. Theglobal fluxes are also mutually consistent withinabout 710% yielding 2.270.2 Pg C yr�1 for the1995 global ocean flux. This suggests that theglobal pCO2 fields estimated using these threedatabase sizes are consistent with each other towithin 710%. The global sea–air CO2 flux isconsistent with the estimate of 2.070.6 Pg C yr�1

obtained on the basis of the changes of theatmospheric oxygen and CO2 concentrations ob-served during the 1990s (Keeling et al., 1996;Battle et al., 2000). A number of model studies(e.g., Sarmiento et al., 1992; Siegenthaler andSarmiento, 1993; Joos and Bruno, 1998) yield amean global ocean uptake of about 2 Pg C yr�1.

The flux estimates presented in Table 1 aresupported by the results from these two indepen-dent lines of investigations.Fig. 2 shows the distribution of the mean annual

net sea–air CO2 flux. The eastern equatorial Pacificand the northwestern Arabian Sea are the mostintense CO2 source areas (red–orange). Thetropical Atlantic and Indian Oceans and thenorthwestern subarctic Pacific are also prominentsource areas. Strong sink areas are found in thetransition zone between the subtropical gyre andsubpolar waters, i.e. 401N–601N and 401S–601S. Inthese areas, low pCO2 waters are formed by thejuxtaposition of the cooling of warm waters withthe biological drawdown of pCO2 in the nutrient-rich subpolar waters. High wind speeds over theselow pCO2 waters increase the CO2 uptake rate byocean waters. The regional flux (%) values listed

30˚

30˚

60˚

60˚

90˚

90˚

120˚

120˚

150˚

150˚

180˚

180˚

150˚

150˚

120˚

120˚

90˚

90˚

60˚

60˚

30˚

30˚

80˚ 80˚70˚ 70˚

60˚ 60˚50˚ 50˚

40˚ 40˚

30˚ 30˚

20˚ 20˚

10˚ 10˚

0˚ 0˚

10˚ 10˚

20˚ 20˚

30˚ 30˚

40˚ 40˚50˚ 50˚

60˚ 60˚70˚ 70˚

80˚ 80˚

-11 -6 -5 -4 -3 -2 -1 0 1 3 5 13

Net Flux (moles CO2 m-2 year-1)

Mean Annual Air-Sea Flux for 1995 (NCEP 41-Yr Wind, 940K, W-92)

Fig. 2. Mean annual net air–sea flux for CO2 (mole CO2m�2 yr�1) for 1995. The following information has been used; (a)

climatological distribution of surface-water pCO2 for the reference year 1995, (b) the NCEP/NCAR 41-year mean wind speeds, (c) the

long-term wind-speed dependence of the sea–air CO2 transfer velocity by Wanninkhof (1992), (d) the concentration of atmospheric

CO2 in dry air in 1995 (GLOBALVIEW-CO2, 2000), and (e) the climatological barometric pressure and sea-surface temperature (Atlas

of Surface Marine Data, 1994). Red–yellow areas indicate that the ocean is a source for atmospheric CO2, and blue–purple areas

indicate that the ocean is a CO2 sink.

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–16221608

near the bottom of Table 1 indicate that, althoughthe Southern Ocean (south of 501S) occupies only10% of the global ocean area, it takes up about20+% of the global ocean CO2 uptake flux.Similarly, the Atlantic Ocean (north of 501S)occupies about 24% of the ocean area and takesup about 40+% of the global ocean CO2 uptake.In contrast, the Pacific Ocean (north of 501S) takesup only 18% of the global ocean CO2 uptake,whereas it occupies 49% of the global ocean area.This is due primarily to the fact that the equatorialPacific is an intense CO2 source area, andsecondarily to the fact that the southeastern SouthPacific is a weak CO2 source area unlike the otheroceans. If the CO2 source flux from the equatorialPacific is reduced during El Ni *no events (Feelyet al., 1999), then the Pacific as a whole wouldbecome a stronger CO2 sink.

3.3. Effects of wind speed and gas transfer velocity

on the CO2 flux

The CO2 flux thus computed depends sensitivelyon the choice of the wind speed field and the wind-speed dependence of the CO2 gas transfer velocityacross the sea–air interface. We consider two setsof wind-speed distributions: the climatologicalmean wind-speed field based on the NCEP/NCARdata collected over the past 41 years, and theNCEP/NCAR annual wind-speed distributionduring 1995. These wind data have been re-sampled for our 41� 51 grid and used to computemean monthly values for each pixel area. For theCO2 gas transfer velocity, we use the (wind speed)

2

dependence of Wanninkhof (1992) (his Eq. (1) fora long-term average wind, named W92) and the(wind speed)3 dependence of Wanninkhof andMcGillis (1999) (their Eq. (5) for a long-termaverage wind, named WM99). Their Rayleighwind-speed distribution function, which is asolution to the Weibull distribution for isotropicwinds, is assumed for both of these equations. Theregional and global sea–air CO2 fluxes computedusing four combinations of the transfer velocityformulations and wind-speed fields are summar-ized in Table 2.The fluxes computed using the W92 and the

NCEP/NACR 41-year mean wind (named 41-yr)

are listed in the first row for each grouping forzonal bands, Oceanic Regions and Regional Flux(%) in Table 2. The column ‘‘Errors in Flux’’located at the right extreme of the table lists thedeviations from the mean flux that have beencomputed using one standard deviation in windspeeds (about 72m s�1 on the global average)above and below from the mean monthly windspeed in each pixel area. These changes in windspeeds affect the regional and global flux values byabout 725%. The fluxes computed using thesingle year mean wind-speed data for 1995 arelisted in the second line in each zonal band groupin the table. The wind speeds for 1995 are muchlower than the 41-year mean in the northernhemisphere and higher over the Southern Ocean.Accordingly, the northern ocean sink fluxes areless than the climatological mean and the SouthernOcean sink fluxes are stronger. The global meanocean uptake flux of �1.8 Pg C yr�1 is about 18%below the climatological mean of �2.2 Pg C yr�1

and is within 720% estimated on the basis of thestandard deviation of the wind-speed data.When the (wind speed)3 dependence is used, the

CO2 fluxes in higher-latitude areas with strongwinds are increased significantly, as are the errorsdue to wind-speed variability. The global oceanuptake flux computed with the 41-year mean wind-speed data and the 1995 data is, respectively, �3.7and �3.0 Pg C yr�1, an increase by about 70%over the flux that is computed using the (windspeed)2 dependence. These flux values are signifi-cantly greater than the flux of �2.070.6 Pg C yr�1

during the 1990s (Keeling et al., 1996; Battle et al.,2000) estimated by the observed changes of theatmospheric CO2 and oxygen concentrations.However, the relative magnitudes for CO2 uptakeby ocean basins (shown in % in the last four rowsin the table) remain nearly unaffected by thechoice of the wind-speed dependence of the gas-transfer velocity.As noted by Wanninkhof et al. (2002), the cubic

parameterization for long-term winds used in thisstudy assumes that the regional wind-speedspectrum follows the Rayleigh distribution func-tion. If the regional distributions are deter-mined from the NCEP 6-h wind products, theglobal flux will decrease to �3.0 Pg C yr�1 for the

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–1622 1609

NCEP-41-year wind and �2.3 Pg C yr�1 for theNCEP-1995 wind data. This indicates that theestimated flux values are also sensitive to wind-speed spectra.

4. Seasonal variation of pCO2 in surface waters

On a global scale, the temperature effect onsurface-water pCO2 is similar in magnitude but

Table 2

The effect of wind speeds and that of the wind-speed dependence of CO2 gas transfer velocity on the net sea–air CO2 flux are shown

using the climatological sea–air pCO2 difference obtained in this work

Lat. band Gas trans./

wind data

Pacific Ocean

(Pg Cyr�1)

Atlantic

Ocean

(Pg Cyr�1)

Indian Ocean

(Pg Cyr�1)

Southern

Ocean

(Pg Cyr�1)

Global

Ocean

(Pg Cyr�1)

Errors in

flux (%)

N of 501N W92/41-yr +0.01 �0.40 F F �0.39 +28, �23W92/1995 +0.03 �0.18 F F �0.14WM99/41-yr +0.03 �0.55 F F �0.52 +44, �35WM99/1995 +0.07 �0.17 F F �0.10

141N-501N W92/41-yr �0.64 �0.34 +0.07 F �0.92 +25, �23W92/1995 �0.29 �0.28 +0.03 F �0.54WM99/41-yr �0.94 �0.48 +0.10 F �1.31 +43, �32WM99/1995 �0.29 �0.38 +0.02 F �0.64

141N-141S W92/41-yr +0.74 +0.15 +0.18 F +1.07 +29, �24W92/1995 +0.61 +0.07 +0.15 F +0.83

WM99/41-yr +0.67 +0.14 +0.20 F +1.00 +43, �31WM99/1995 +0.62 +0.05 +0.12 F +0.79

141S-501S W92/41-yr �0.51 �0.33 �0.67 F �1.51 +22, �20W92/1995 �0.57 �0.31 �0.50 F �1.38WM99/41-yr �0.68 �0.51 �0.97 F �2.16 +37, �30WM99/1995 �0.88 �0.51 �0.63 F �2.02

S of 501S W92/41-yr F F F �0.47 �0.47 +26,�21W92/1995 F F F �0.58 �0.58WM99/41-yr F F F �0.74 �0.74 +41, �32WM99/1995 F F F �1.02 �1.02F F F F F F F

Oceanic W92/41-yr �0.40 �0.92 �0.43 �0.47 �2.22 +22, �19Regions W92/1995 �0.21 �0.69 �0.33 �0.58 �1.81

WM99/41-yr �0.92 �1.39 �0.67 �0.74 �3.72 +40, �32WM99/1995 �0.48 �1.01 �0.48 �1.02 �3.00

Regional W92/41-yr 18 41 19 21 100

Flux (%) W92/1995 12 38 18 32 100

WM99/41-yr 25 37 18 20 100

WM99/1995 16 34 16 34 100

The flux values have been computed using the (wind speed)2 dependence of CO2 gas transfer velocity by Wanninkhof (1992; named

W92) and the (wind speed)3 dependence by Wanninkhof and McGillis (1999; named WM99), respectively, for each of the two sets of

wind data: the NCEP/NCAR 41-year (named 41-yr in the table) and 1995 mean monthly wind speeds. Errors in flux (% in the flux)

listed in the extreme right column represent the flux changes resulting from + or � one standard deviation (about 72m s�1 on the

global average) from the annual mean wind speed in each pixel area. The positive errors in the flux represent when the mean monthly

wind speed over each pixel area was increased by one standard deviation; and the negative errors represent when the mean wind speed

was reduced by one standard deviation

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–16221610

opposite in direction to the biological effect, inwhich changes of the total CO2 concentration isthe dominant factor. The pCO2 in surface oceanwaters doubles for every 161C temperatureincrease (qln pCO2/qT=0.04231C�1, Takahashiet al., 1993). For a parcel of seawater withconstant chemical composition, its pCO2 wouldbe increased by a factor of 4 when it is warmedfrom polar water temperatures of about �1.91C toequatorial water temperatures of about 301C. Onthe other hand, the total CO2 concentration insurface waters ranges from about 2150 mmol kg�1

in polar waters to 1850 mmol kg�1 in tropicalwaters. If a global mean Revelle factor(qln pCO2/qln TCO2) of 10 is used, this decreaseof TCO2 should cause a reduction of surface-waterpCO2 by a factor of 1/4.5 (=(pCO2)tropical/(pCO2)polar=(1850/2150)10). From the climatolo-gical mean monthly pCO2 data, we can derive thedistribution and magnitude of seasonal amplitudeof surface-water pCO2 over the global oceans. Theeffect of temperature on seasonal amplitude isdifferentiated from that of the TCO2 effect, andthe relative importance of these processes forregulating the seasonal amplitude in pCO2 ismapped over the global oceans. This informationis important not only for understanding the time–space variability for the sea–air CO2 exchange, butalso for planning future field investigations.Although the maximum level of pCO2 and theamount of nutrients available for the photosynth-esis are governed primarily by the vertical flux ofsubsurface waters, this flux is difficult to measuredue to the lack of widely distributed tracers, andcan only be estimated reliably in specific areas suchas the Weddell Sea (e.g., Martinson, 1990;Martinson and Iannuzzi, 1998). Therefore, theeffect of the vertical flux of deep waters is notaddressed in this paper.

4.1. Method of analysis and examples

The method used for the analysis will beillustrated using seasonal data observed at threecontrasting locations.

4.1.1. Seasonal pCO2 variation at the BATS site,

North Atlantic

Fig. 3A shows the multi-year observations from1996 through early 1998 for the water temperatureand pCO2 made in mixed-layer waters in thevicinity of the BATS site (321500N and 641100W)near Bermuda (Bates, 2001). The temperature(open circles) and pCO2 (filled circles) are lowest

Year1996 1997 1998 1999

Sea

wat

erpC

O2

atS

ST

(at

m)

250

275

300

325

350

375

400

425

Sea

Sur

face

Tem

per

atur

e(

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34pCO2

at SST

SST

Mean

SST

Mean

pCO2

Year1996 1997 1998 1999

Sea

wat

erpC

O2

(at

m)

275

300

325

350

375

400

425

450

Mean pCO2 corrected for SST

pCO2

at 23.4 C

(a)

(b)

Fig. 3. (A) The surface-water pCO2 (filled circles) and the

temperature of mixed-layer water (SST, open circles) observed

at and near the BATS station in 1996 through early 1998. (B)

The pCO2 values at the mean water temperature of 23.41C

(open circles) and the annual mean pCO2 value corrected for

changes in temperature (filled circles).

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–1622 1611

in winter (February) and highest in summer (July),and their seasonal changes are closely in phase,suggesting that the change in pCO2 is primarilygoverned by the temperature change. The peak-to-peak amplitude of seasonal variation in tempera-ture and pCO2 is 9.5 (70.3)1C and 100 (75) matm,respectively, in 1996 and 1997. In order to removethe temperature effect from the observed pCO2,the observed pCO2 values are normalized to aconstant temperature of 23.41C, the mean annualtemperature of seawater at the station, using theequation

ðpCO2 at TmeanÞ ¼ ðpCO2Þobs� exp½0:0423ðTmean � TobsÞ; ð1Þ

where T is the temperature in 1C, and thesubscripts ‘‘mean’’ and ‘‘obs’’ indicate the annualaverage and observed values, respectively.Changes in this quantity represent primarilychanges in the total CO2 concentration. Thetemperature effect on pCO2 for isochemical sea-water (qln pCO2/qT) of 0.04231C�1 was deter-mined experimentally by Takahashi et al. (1993)for a North Atlantic surface-water sample. This isnearly independent of temperature and chemicalcomposition (i.e. alkalinity/total CO2 ratio) ofseawater (Rubin, 2000), and may be consideredconstant. As shown with the open circles inFig. 3B, the constant-temperature pCO2 valuesdecrease by about 55 matm from winter throughsummer in each of 1996 and 1997. This decreaseapproximates the ‘‘net biology’’ effect, whichincludes the effects of the net CO2 utilization, asmall amount of net alkalinity change due tocarbonate production and nitrate utilization, sea–air exchange of CO2, and an addition of CO2 andalkalinity by the vertical mixing of subsurfacewaters. However, from winter to summer, themixed-layer depth tends to become shallower withprogressing season, and hence the vertical trans-port of subsurface waters is considered small. Thesea–air gas flux during this period is considered tobe small since the sea–air pCO2 difference changesfrom positive to negative about midway during thewarming period and has a similar magnitude.Bates (2001) observed that, on the average, thetotal CO2 concentration in mixed-layer watersdecreased from winter to summer by about

33 mmol kg�1 (normalized to a constant salinityof 36.6) in both of these years, while the salinity-normalized alkalinity values stayed nearly con-stant. Using a Revelle factor of 10, the observedseasonal decrease in TCO2 can be translated to apCO2 change of about 60 matm. Hence, consider-ing minor changes in the alkalinity, the observedwinter-to-summer drawdown of TCO2 is consis-tent with the net biological effect estimated usingthe constant-temperature pCO2 values.The effect of temperature changes on pCO2 has

been computed by perturbing the mean annualpCO2 of 346 matm with the difference between themean and observed temperature. The pCO2 valueat a given observed temperature, Tobs; has beencomputed using the equation

ðpCO2 at TobsÞ ¼ ðMean annual pCO2Þ

exp½0:0423ðTobs � TmeanÞ; ð2Þ

where the mean annual temperature, Tmean; of23.41C is used. The resulting values, that areshown in Fig. 3B with filled circles, indicate thepCO2 values that would be expected only fromtemperature changes, if a parcel of water havingthe mean pCO2 value of 346 matm is subjected toseasonal temperature changes under isochemicalconditions. A mean seasonal amplitude of about150 matm is observed in 1996 and 1997. Thissuggests that the increasing effect of winter-to-summer warming on the seawater pCO2

(150 matm) is partially compensated by the low-ering effect of biological utilization of CO2

(55 matm) to yield the observed seasonal pCO2

amplitude of 100 matm. The relative importance ofthe temperature and biological effects may beexpressed in terms of the ratio, (temperatureeffect)/(biology effect) or difference [(temperatureeffect)�(biological effect)]. At this station, theratio is estimated to be 2.7 (=150/55 matm), andthe difference is +95 matm.

4.1.2. Seasonal pCO2 variation in the Ross Sea,

Antarctica

Fig. 4 shows the seasonal variation of surface-water pCO2 observed during the US JGOFSprogram in 1996–1998, along the 76.51S latitudebetween 1691E and 1771W in the southern RossSea, Antarctica. The observations were made from

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–16221612

November 1996, through the end of February1997, when the area was not covered with ice(Sweeney, 2000). The seasonal variation is ap-proximated by a sine curve in Fig. 4, showing thatthe seawater pCO2 decreases by about 260 matmfrom about 410 matm in early November to about150 matm in mid-January. Here, the mixed-layerwater temperature is nearly constant rangingbetween –1.01C and –1.51C throughout the openwater period, and the effect of temperature onpCO2 is small (about 5 matm). The seasonalvariation of pCO2 is entirely attributed to thebiological utilization of CO2. Accordingly, the(temperature effect)/(biology effect) ratio is 0.02(=5/260) or the difference is �255 matm.

4.1.3. Seasonal pCO2 variation at the Weather

Station P, North Pacific

Fig. 5A shows seasonal variations in pCO2 andtemperature observed in mixed-layer waters byWong and Chan (1991) at the Weather Station‘‘P’’ in the eastern subarctic Pacific (501N, 1451W)during 1973–1978. Since their observations havedata gaps, the data collected over the 5-year periodare plotted for a single year to give a composite

seasonal cycle. While the temperature changedseasonally from about 4.51C to 131C in asinusoidal pattern with a spring minimum and asummer maximum, the observed pCO2 values at insitu temperatures exhibit irregular variations

Fig. 4. Seasonal changes in the surface-water pCO2 observed

along 76.51S in the Ross Sea, Antarctica. Julian date 0

represents January 1, 1997. The solid curve is a trend line

approximated by a simple sine function.

Julian Date0 30 60 90 120 150 180 210 240 270 300 330 360

Sea

wat

erpC

O2

(at

m)

230

240

250

260

270

280

290

300

310

320

330

340

350

360

370

380

390

400

Sea

Sur

face

Tem

pera

ture

(

4

5

6

7

8

9

10

11

12

13

14

15

pCO2

at SST

SST

Mean

pCO2

Mean

SST

Julian Date0 30 60 90 120 150 180 210 240 270 300 330 360

Sea

wat

erpC

O2

(at

m)

225

250

275

300

325

350

375

400

NO

2(

mol

kg-1

)

0

5

10

15

20

25

30

35

Mean pCO2 corrected for SSTpCO2

at 8.34

NO3

(a)

(b)

Fig. 5. (A) The pCO2 and water temperature observed in the

mixed layer observed in 1972–1975 at Weather Station ‘‘P’’ by

Wong and Chan (1991) are shown with filled and open circles

respectively. The multi-year observations are plotted in a single

year in order to show a complete annual cycle. (B) The pCO2

values normalized to the mean annual temperature of 8.341C

are shown with open circles, and the mean annual pCO2 values

corrected for temperature changes are shown with filled circles.

The concentrations of nitrate observed at this station by

Parslow (1981) are shown with ‘‘x’’.

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–1622 1613

(three maxima and three minima in a year)fluctuating between 290 and 340 matm. The pCO2

values observed at this station exhibit not onlymuch smaller seasonal amplitudes, but also morecomplex patterns compared to those observed atthe BATS station (Fig. 3A). In order to differ-entiate the effect of biology from that of tempera-ture change, we have computed the pCO2 values ata mean annual temperature of 8.341C (opencircles, Fig. 5B) using Eq. (1). The biological effectestimated from the temperature normalization isabout 115 matm (375 matm in February and260 matm in July–August, Fig. 5B), twice as largeas that observed near Bermuda. At this location,Parslow (1981) has observed a nitrate drawdownof about 8 mmol kg�1 (see the ‘‘x’’ symbols inFig. 3B) from March through July. Assuming theRedfield C/N ratio of 106/15 and the Revellefactor (qln pCO2/qln TCO2) of 12, this nitratedrawdown corresponds to a TCO2 drawdown of56.5 mmol kg�1 and a pCO2 decrease of 115 matm.Thus the magnitude of the biological effectestimated on the basis of the constant-temperaturepCO2 values is consistent with the net biologicalutilization of CO2 estimated on the basis of theseasonal drawdown of nitrate. Since the biologicaldrawdown of pCO2 occurs mostly during theperiod when the mixed layer is getting shallowerwith progressing season, the vertical transportcontribution of CO2-rich subsurface waters isconsidered minimal.Fig. 5B shows that the seasonal warming of

water causes pCO2 to increase by about 105 matmfrom about 270 matm in February to 375 matm inJuly–August. Thus, the temperature effect issimilar in magnitude to the biological effect, butthey are approximately 6 months out of phase asphotosynthesis is increased at warmer tempera-tures (Fig. 5B). Hence, at this station, theypartially cancel each other, so that the seasonalvariation of surface-water pCO2 is reduced inamplitude without simple sinusoidal seasonalvariations. The temperature/biology effect ratioand difference in the effects at this site are,respectively, about 0.9 (=10/115 matm) and�10 matm.Using the method described above, global-ocean

maps showing the respective effects of temperature

and biology on the surface-water pCO2 and theratio or difference of these two competing effectsmay be constructed. The effect of biology on thesurface-water pCO2 in a given pixel area,ðDpCO2Þbio; is represented by the seasonal ampli-tude of pCO2 values corrected to the mean annualtemperature in a given area, (pCO2 at Tmean), usingEq. (1)

ðDpCO2Þbio ¼ ðpCO2 at TmeanÞmax� ðpCO2 at TmeanÞmin; ð3Þ

where the subscripts ‘‘max’’ and ‘‘min’’ indicatethe seasonal maximum and minimum values.The effect of temperature changes on the mean

annual pCO2 value, (DpCO2)temp, is represented bythe seasonal amplitude of (pCO2 at Tobs) valuescomputed using Eq. (2)

ðDpCO2Þtemp ¼ ðpCO2 at TobsÞmax� ðpCO2 at TobsÞmin: ð4Þ

The relative importance of the biology andtemperature effects is expressed by the ratio, T=B;or the difference ðT � BÞ:

ðT=BÞ ¼ ðDpCO2Þtemp=ðDpCO2Þbio or

ðT � BÞ ¼ ðDpCO2Þtemp � ðDpCO2Þbio: ð5Þ

In oceanic areas where the effect of temperaturechanges on surface-water pCO2 exceeds thebiological effect (e.g., BATS site), the ðT=BÞ ratiois greater than 1 or ðT � BÞ positive, whereas inareas where the biology effect exceeds the tem-perature effect, the ðT=BÞ ratio varies from 0 to1or ðT � BÞ is negative. The two effects cancel eachother in areas where the ðT=BÞ ratio is 1 or ðT � BÞis 0.

4.2. Global distribution of seasonal pCO2 amplitude

The magnitude of seasonal amplitude has beencomputed by taking the difference between themaximum and minimum mean monthly surface-water pCO2 values in each pixel area (Fig. 6). Inorder to express phase differences in seasonalchanges, the following sign conventions are used.When a pCO2 maximum is found during theperiod for SST below the annual mean (i.e. colderseasons, or winter), the seasonal amplitude for

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–16221614

pCO2 is expressed by (minimum–maximum), anegative value. On the other hand, when a pCO2

maximum is found during the warmer seasons (orsummer), the amplitude is expressed by (max-imum–minimum), a positive value. This conven-tion has an advantage that differences in timing ofseasonal changes within each hemisphere aredepicted by the positive (red–yellow) and negative(blue–purple) values. According to this signconvention, the magnitude of amplitude (andcolors) observed in the northern oceans can becompared with that (or the same colors) for thesouthern oceans. However, they are, in fact, about6 months out of phase.In Fig. 6, seasonal amplitudes as negative as

�215 matm (blue–purple) are observed in subpolar

and polar regions. This is a result of the biologicaldrawdown of pCO2 during summer and theincrease in pCO2 during winter due to the verticalmixing of high-CO2 subsurface waters. Thepositive values (yellow and red) are found in themid-latitude oceans where the warm period pCO2

values are greater than the cold period values.Seasonal amplitudes as large as 120 matm(282 matm in March and 403 matm in August) areobserved in the mid-latitude areas. These are theareas where seasonal temperature changes accountfor more than 50% of the seasonal pCO2 changes.The Arabian Sea is an exception to the mid-latitude seasonality, and seasonal amplitudes aslarge as 174 matm are observed. Here, the mini-mum monthly pCO2 value of 355 matm is observed

GMT 2001 Jul 10 16:50:33

40˚

40˚

80˚

80˚

120˚

120˚

160˚

160˚

160˚

160˚

120˚

120˚

80˚

80˚

40˚

40˚

80˚ 80˚70˚ 70˚

60˚ 60˚50˚ 50˚

40˚ 40˚

30˚ 30˚

20˚ 20˚

10˚ 10˚

0˚ 0˚

10˚ 10˚

20˚ 20˚

30˚ 30˚

40˚ 40˚50˚ 50˚

60˚ 60˚70˚ 70˚

80˚ 80˚

-215 -195 -80 -60 -40 -20 0 20 40 60 80 140 175

pCO2 Difference (µatm)

Seasonal Mean Monthly Amplitude of pCO2 in Seawater, (1995, W-92)

Fig. 6. Distribution of the peak-to-peak seasonal amplitude of surface-water pCO2. The positive values indicate areas where the

surface-water pCO2 maximum occurs during warm-water seasons, and the negative values indicate areas where pCO2 maximum occurs

during cold-water seasons. Within each hemisphere, seasonal changes for the temperate oceans (positive values) are approximately 6

months out of phase from those (negative values) for the subpolar and polar oceans. The seasonal variation in the northern hemisphere

is about 6 months out of phase from that in the southern hemisphere. In several pixels in the subarctic Atlantic, eastern equatorial

Pacific, Arabian Sea, Chilean coast, and Antarctic coast, abrupt changes in the signs and colors (from red to dark blue) are observed.

Because of the coarse spatial resolution of our analysis, the pCO2-SST relations become reversed in areas with fine-scale oceanographic

features such as the East Greenland Current in the subarctic Atlantic, the Brazil Current in the South Atlantic and coastal upwelling

areas.

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–1622 1615

during March; and the maximum monthly pCO2

of 529 matm occurs in July, during which amaximum daily value as high as 640 matm isobserved. The extremely high values are dueprimarily to the monsoon-induced upwelling ofdeep waters. Large seasonal amplitudes are alsoobserved near the Chilean coast, where strongupwelling is known to occur. The green areaswhere seasonal changes are small are found alongthe boundary between the warm subtropicalwaters and the colder subpolar waters of theSouthern Ocean. This is due to the near cancella-tion of the temperature effect by the biology effect.

4.3. Distribution of the biological effects on

surface-water pCO2

The global distribution of the biological effecton surface-water pCO2 (Fig. 7) has been calculated

using Eq. (3). Very large biological drawdowneffects on pCO2 exceeding 140 matm (yellow–red)are observed in high-latitude northern oceansnorth of about 401N (yellow–orange–red).The effects are especially intense in thenorthwestern subarctic Pacific, where theeffects as intense as 264 matm are observed.The subtropical and tropical oceans in bothhemispheres exhibit a small biologicaldrawdown smaller than 50 matm (light–darkblue). The eastern equatorial Pacific, however,shows intense biological effects up to 276 matm.Moderate to large biological effects rangingfrom 75 to 120 matm are seen in the southernhemisphere oceans between 301S and 501S. Theeffect tends to decrease to higher-latitude areas inthe Southern Oceans, although locally strongdrawdown areas (such as the Ross Sea) areobserved.

GMT 2001 Jul 10 16:50:44

40˚

40˚

80˚

80˚

120˚

120˚

160˚

160˚

160˚

160˚

120˚

120˚

80˚

80˚

40˚

40˚

80˚ 80˚70˚ 70˚

60˚ 60˚50˚ 50˚

40˚ 40˚

30˚ 30˚

20˚ 20˚

10˚ 10˚

0˚ 0˚

10˚ 10˚

20˚ 20˚

30˚ 30˚

40˚ 40˚50˚ 50˚

60˚ 60˚70˚ 70˚

80˚ 80˚

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 280

pCO2 Drawdown (µatm)

Seasonal Biological Drawdown of Seawater pCO2

Fig. 7. The effect of the biological utilization of CO2 is represented by the seasonal amplitude for the pCO2 values corrected to the

mean water temperature for each pixel area using Eq. (3). Values exceeding 150 matm (yellow–orange–red) are observed in the

northwestern subarctic Pacific and Atlantic, the eastern equatorial Pacific, the northwestern Arabian Sea, and the Ross Sea. These high

biology areas are associated with intense upwelling of subsurface waters with high CO2 and nutrient concentrations. Small areas near

the coasts of Mexico, Chile and Argentine also show strong biological drawdown effects. We have, however, insufficient seasonal data

to characterize the seasonal changes in the southern Weddell Sea and the coastal upwelling areas off West Africa.

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–16221616

We observe that the oligotrophic subtropicalgyre areas between 301N and 301S (but excludingthe 101N–101S equatorial Pacific belt and theupwelling areas located in the eastern tropicalPacific and Atlantic and in the Arabian Sea) havean average biological pCO2 drawdown of about35 matm (Fig. 7), whereas the concentrations ofnutrients are very low throughout the year withnegligibly small seasonal changes. The pCO2

drawdown corresponds to a TCO2 decrease ofabout 20 mmol kg�1 that occurred between winterand early summer time when the pCO2 in seawaterbecomes minimum. The sea–air CO2 flux over theoligotrophic waters during the growth period issmall as seen in Fig. 2 and may be neglected. Usinga mean mixed-layer depth of 50m (World OceanDatabase, 1998), the net biological CO2 utilizationin the mixed layer is estimated to be about 1moleCm�2 for the growth period during which water iswarming and mixed-layer depth becomes shal-lower with the progressing of the season fromspring to summer. This is a minimum estimatesince the effect of TCO2 addition by upwelling isneglected. Since this represents early growths overthe first 6 months in a year, the annual productionmay be close to 2mole Cm�2 yr�1. This may becompared with the annual export production forthe global oligotrophic ocean estimated by Lawset al. (2000) based upon the SeaWiFS ocean-colordata, the vertically generalized production model(VGPM) of Behrenfeld and Falkowski (1997) anda pelagic food web model. Their PTE (primaryproduction and temperature) model solutions givea mean export production (which is assumed to beequal to the new production) of about 20 gCm�2 yr�1 (or 1.7mole Cm�2 yr�1) for theoligotrophic oceans in both hemispheres. Thus,these net production values estimated usingindependent methods are consistent with eachother.

4.4. Distribution of the temperature effect on pCO2

The effect of seasonal temperature changes onsurface-water pCO2 is calculated using Eq. (4) andthe World Ocean Database (1998) SST data recastonto our 41� 51 grid. The seasonal amplitude isshown in Fig. 8 to depict the global distribution of

the effect on surface-water pCO2 of seasonalchanges in SST. The tropical oceans between201N and 201S have small seasonal temperaturechanges, and hence small temperature effects onpCO2 (purple areas). The eastern equatorial areasof the Pacific and Atlantic Oceans are exceptionsdue to seasonal upwelling of deep waters. Thesouthern high-latitude areas south of about 551Salso show small temperature effects. Large tem-perature effects exceeding 80 matm are observed inthe mid-latitude areas in both hemispheres. Thelargest temperature effects, as large as 217 matm,are observed in the confluence areas of the warmKuroshio with the cold Oyashio current waters inthe northwestern Pacific and of the warm GulfStream with the cold Labrador Current waters inthe northwestern Atlantic. As shown in Fig. 7,these areas coincide with strong biological effectareas. Since the temperate and biological effectsare out of phase in these areas, they tend to canceleach other. However, in the northwestern Pacific,the biology effect exceeds the temperature effect,whereas in the northwestern Atlantic, the tem-perature effect exceeds the biology effect (seeFig. 9).

4.5. Relative importance of the temperature and

biology effects

Fig. 9 shows the global distribution of thedifference between the temperature and biologyeffects (T � B) on surface-water pCO2 computedusing Eq. (5). The areas dominated by the biologyeffect are indicated with green–blue (negativevalues); and those dominated by the temperatureeffect are indicated with yellow–orange (positivevalues). In general, the biology effect dominatesthe temperature effects in high-latitude oceans inboth hemispheres. The equatorial oceans are alsodominated by the biology effect. In temperateoceans of both hemispheres, the temperature effectexceeds the biology effect. Along the polewardedge and equatorward edge of the subtropicalregion, the ðT � BÞ becomes nearly zero (green–yellow boundary). This indicates that the effectof temperature cancels with the effect of tempera-ture, and hence that seasonal amplitude of

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–1622 1617

surface-water pCO2 is small along the transitionareas between yellow and green.Large areas between 401S and 701S in the

Southern Ocean have negative values (green)indicating that the biology effect exceeds thetemperature effect. This feature is demonstratedby the extensive pCO2 data obtained between 401Sand 651S during different seasons in the Atlantic,Indian and Pacific sectors of the Southern Oceanunder the French KERFIX/F-JGOFS, AustralianWOCE and US AESOPS and WOCE programs.The biology effect intensifies to the south, espe-cially in the areas south of the polar front (locatedabout 601S) and the shelf areas of Antarctica. Thissuggests that the photosynthetic drawdown of CO2

in the open Southern Ocean waters, as well as theshelf waters around Antarctica, plays an impor-tant role for the transport of atmospheric CO2 intothe deep ocean regime via the thermohalinecirculation process.

5. Summary and conclusions

On the basis of about 940,000 surface-waterpCO2 observations, monthly maps showing theclimatological distribution and seasonal amplitudefor surface-water pCO2 over the global oceanrepresenting non-El Ni *no conditions have beenconstructed for a reference year 1995 with a 41� 51spatial resolution. The net sea–air CO2 flux overthe regional and global sea–air ocean has beencomputed using the NCEP/NCAR 41-year meanmonthly wind speeds and the (wind speed)2

dependence of sea–air CO2 gas transfer velocityby Wanninkhof (1992) for long-term wind. A netCO2 uptake flux by the global oceans of 2.2 PgC yr�1 is obtained. This uptake flux is consistentwithin 70.2 Pg Cyr�1 with the distribution of thesea–air pCO2 difference obtained previously using250,000 (Takahashi et al., 1997) and 550,000(Takahashi et al., 1999) measurements. Based

GMT 2001 Jul 10 16:50:51

40˚

40˚

80˚

80˚

120˚

120˚

160˚

160˚

160˚

160˚

120˚

120˚

80˚

80˚

40˚

40˚

80˚ 80˚70˚ 70˚

60˚ 60˚50˚ 50˚

40˚ 40˚

30˚ 30˚

20˚ 20˚

10˚ 10˚

0˚ 0˚

10˚ 10˚

20˚ 20˚

30˚ 30˚

40˚ 40˚50˚ 50˚

60˚ 60˚70˚ 70˚

80˚ 80˚

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 280

pCO2 Change (µatm)

Seasonal Temperature Effect on Seawater pCO2

Fig. 8. The effect of seasonal temperature changes on surface-water pCO2. This is represented by the seasonal amplitude of the mean

annual pCO2 values corrected for seasonal water temperature changes in each pixel area using Eq. (4).

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–16221618

upon one standard deviation of mean monthlywind speeds over each pixel area, the errorsassociated with wind-speed variability have beenestimated to be about 720% of the mean annualflux. This estimate for the global-ocean uptake fluxis consistent with the values of 2.070.6 Pg C yr�1

estimated on the basis of the observed changes inthe atmospheric CO2 and oxygen concentrationsduring the 1990s (Keeling et al., 1996; Battle et al.,2000). The CO2 flux obtained by the present studydepends on the relationship between wind speedon sea–air CO2 gas transfer velocity used for theflux computation. If the (wind speed)3 dependenceof Wanninkhof and McGillis (1999) for long-termwind is chosen, the annual ocean uptake as well asthe sensitivity to wind variability would beincreased by about 70%. The global-ocean uptakeflux of 3.7 Pg C yr�1 estimated using the (windspeed)3 relation with a Rayleigh wind spectrum issignificantly greater than the fluxes estimated fromthe atmospheric CO2 and oxygen data.

An ocean zone between 401 and 601 in latitudesin both northern and southern hemispheres is amajor sink for atmospheric CO2 (see Fig. 2). Inthese areas, poleward-flowing warm waters are incontact and mix with the subpolar waters rich innutrients. The pCO2 in surface waters is decreasedby the juxtaposition of the cooling effect of warmwaters with the biological drawdown effect onpCO2 in subpolar waters. High wind speeds overthese low pCO2 waters increase the CO2 uptakerate by ocean waters.The seasonal amplitude of surface-water pCO2

at a given location is regulated by the biologicalutilization of CO2 and temperature, while theseasonal maximum pCO2 is governed by theupward flux of high-CO2 subsurface waters andtemperature. In this paper, only the effects of thenet biological utilization and seasonal temperaturechanges on surface-water pCO2 are discussed. Theseasonal amplitude (as defined by the differencebetween the maximum and minimum values in

GMT 2001 Jul 10 16:50:26

40˚

40˚

80˚

80˚

120˚

120˚

160˚

160˚

160˚

160˚

120˚

120˚

80˚

80˚

40˚

40˚

80˚ 80˚70˚ 70˚

60˚ 60˚50˚ 50˚

40˚ 40˚

30˚ 30˚

20˚ 20˚

10˚ 10˚

0˚ 0˚

10˚ 10˚

20˚ 20˚

30˚ 30˚

40˚ 40˚50˚ 50˚

60˚ 60˚70˚ 70˚

80˚ 80˚

-210 -150 -135 -120 -105 -90 -75 -60 -45 -30 -15 0 15 30 45 60 75

pCO2 (µatm)

Fig. 9. The difference (T � B) between the effects on pCO2 of seasonal temperature changes and biology. The yellow–orange areas

(positive values) are where the temperature effects exceed that of the biological utilization of CO2; the green–blue areas (negative

values) are where the biological effect dominates the seasonal changes in surface-water pCO2.

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–1622 1619

monthly mean pCO2) is as large as +120 matm intemperate oceans and �215 matm in the polaroceans. The sign change indicates that theseasonality in the subpolar and polar areas isapproximately 6 months out of phase from that inthe temperate and tropical ocean areas. Because ofthis phase difference, a zone of near-zero seasonalamplitude is formed when the amplitudes for theseeffects are nearly equal but opposite in theseasonal phase. Such a zone is found along thepoleward and equatorward borders of subtropicalgyres as indicated by the yellow–green colortransition in Fig. 9.Combining the surface-ocean pCO2 data with

the climatological SST data, the biological com-ponent in seasonal pCO2 changes is separatedfrom the temperature change component, and therelative importance of these two effects is pre-sented. The biological component is estimated bythe seasonal amplitude of the temperature-normal-ized pCO2 values. The temperature component ischaracterized by the seasonal amplitude of theannual mean pCO2 value in a given area correctedfor seasonal temperature variation. The globalmaps showing each of these components and theirrelative importance are presented. The seasonalchanges in the following areas are found to bestrongly controlled by the biological processes; theequatorial and subpolar-polar oceans, the north-western Arabian Sea, the eastern equatorialPacific, the upwelling area off the Chile coast,the subarctic northwestern Pacific, the subarcticNorth Atlantic and the coastal waters aroundAntarctica (including the Ross Sea). On the otherhand, in the subtropical gyre areas, the effect oftemperature change exceed the biological effect byas much as 60 matm or a factor of 4. Although thebiological effect is small in these areas, it is about35 matm on the average over the oligotrophicwaters (101N–301N and 101S–301S). This pCO2

effect is equivalent with a biological CO2 utiliza-tion of 20 mmol CO2 kg

�1 for a period from winterthrough the spring growth period, and is broadlyconsistent with the export production estimated byLaws et al. (2000) using the SeaWiFS data.The climatological pCO2 data presented in this

paper show the importance of high-latitude north-ern and southern oceans as a sink for atmospheric

CO2. These areas are also the source areas for thedeep and intermediate water masses and hencerepresent a direct pathway for CO2 exchangebetween the atmosphere and deep oceans. There-fore, the mechanisms that control the sea–air CO2

flux and the CO2 concentration in newly formedwaters must be understood in order to predict thelong term storage of CO2 in deep ocean waters. Weneed to understand the oceanic biogeochemicalfeedback processes in response to the anthropo-genic loading of atmospheric CO2 as well aschanges in the climate and associated changes inocean ecology and biogeochemistry (e.g., Karl,1999). The findings of this study should serve as abasis for characterizing future changes in theoceanic CO2 uptake.The monthly pCO2 values and other values

presented in this paper are available in an electronicformat at a FTP site at the Lamont–Doherty EarthObservatory of Columbia University via an e-mailaddress /[email protected].

Acknowledgements

This study was made possible by internationalcollaboration of field measurements of surface-water pCO2 obtained over many years. We grate-fully acknowledge the financial support offered tous by various government agencies in Australia,France, Iceland, Japan and the United States(National Science Foundation, National Oceano-graphic and Atmospheric Administration andDepartment of Energy). The database construc-tion and data analysis have been supported bygrants from the US National Oceanographic andAtmospheric Administration. We thank for theencouragements and assistance offered by PaulTreguer, Chairman of the Southern Ocean CO2

Symposium held in Brest, France, during thepreparation of this paper. This is ContributionNo. 6226 of LDEO.

References

Atlas of Surface Marine Data, 1994. CD-ROM NODC-56,

Ocean Climate Laboratory, NOAA, Washington, DC.

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–16221620

Bates, N.R., 2001. Interannual variability of oceanic CO2 and

biogeochemical properties in the Western North Atlantic

subtropical gyre. Deep-Sea Research II 48 (8–9), 1507–1528.

Bates, N.R., Takahashi, T., Chipman, D.W., Knapp, A.H.,

1998. Variability of pCO2 on diel to seasonal time scales in

the Sargasseo Sea. Journal of Geophysical Research 103,

15567–15585.

Battle, M., Bender, M.L., Tans, P.P., White, J.W.C., Ellis, J.T.,

Conway, T., Francey, R.J., 2000. Global carbon sinks and

their variability inferred from atmospheric O2 and d13C.Science 287, 2467–2470.

Behrenfeld, M.J., Falkowski, P.G., 1997. Photosynthetic rates

derived from satellite-based chlorophyll concentration.

Limnology and Oceanography 42, 1–20.

Etcheto, J., Boutin, J., Dandonneau, Y., Bakker, D.C.E., Feely,

R.A., Ling, R.D., Nightingale, P.D., Wanninkhof, R., 1999.

Air–sea CO2 flux variability in the equatorial Pacific Ocean

near 1101W. Tellus 51B, 734–747.

Feely, R.A., Wanninkhof, R., Takahashi, T., Tans, P., 1999.

Influence of El Nino on the Equatorial Pacific contribution

to atmospheric CO2 accounts. Nature 398, 597–601.

GLOBALVIEW-CO2: Cooperative Atmospheric Data Integra-

tion ProjectFCarbon Dioxide, 2000. CD-ROM, NOAA

CMDL, Boulder, CO. Also available on Internet via

anonymous FTP to ftp.cmdl.noaa.gov, Path: ccg/co2/

GLOBALVIEW.

Inoue, H.Y., Matsueda, H., Ishii, M., Fushimi, K., Hirota, M.,

Asanuma, I., Takasugi, Y., 1995. Long-term trend of the

partial pressure of carbon dioxide (pCO2) in surface waters

of the western North Pacific 1984–1993. Tellus 47B, 391–

413.

Jenkins, W.J., 1991. Determination of isopycnal diffusivity in

the Sargasse Sea. Journal of Physical Oceanography 21,

1058–1061.

Joos, F., Bruno, M., 1998. Long-term variability of the

terrestrial and oceanic carbon sinks and the budgets of the

carbon isotopes 13C and 14C. Global Biogeochemical Cycles

12, 277–295.

Karl, D.M., 1999. A sea of change: biogeochemical variability

in the North Pacific subtropical gyre. Ecosystems 2, 181–

214.

Keeling, C.D., Whorf, T.P., 2000. Atmospheric CO2 concen-

trationsFMauna Loa Observatory, Hawaii, 1958–1999

(revised August 2000). CDIAC NDP-001 (http://cdiac.es-

d.ornl.gov/ndps/ndp001.html), Carbon Dioxide Informa-

tion and Analysis Center, Oak Ridge, TN.

Keeling, R., Piper, S.C., Heinmann, M., 1996. Global and

hemispheric CO2 sinks deduced from changes in atmo-

spheric O2 concentration. Nature 381, 218–221.

Laws, E.A., Falkowski, P.G., Smith Jr., W.O., Ducklow, H.,

McCarth, J.J., 2000. Temperature effects on export produc-

tion in the open ocean. Global Biogeochemical Cycles 14,

1231–1246.

Martinson, D.G., 1990. Evolution and the southern Antarctic

mixed layer and sea–ice; open ocean deep water formation

and ventilation. Journal of Geophysical Research 95,

11641–11654.

Martinson, D.G., Iannuzzi, R.A., 1998. Antarctic ocean–ice

interaction: implications from ocean bulk property distribu-

tions in the Weddell gyre. Antarctic Sea Ice Physical

Processes, Interactions and Variability, Antarctic Research

Series, Vol. 74. American Geophysical Union, Washington,

DC, pp. 243–271.

Metzl, N., Poisson, A., Louanchi, F., Brunet, C., Schauer, B.,

Bres, B., 1995. Spatio-temporal distributions of air–sea

fluxes of CO2 in the Indian and Antarctic oceans: a first

step. Tellus 47B, 56–69.

Metzl, N., Tilbrook, B., Poisson, A., 1999. The annual fCO2

cycle and the air–sea CO2 flux in the sub-Antarctic Ocean.

Tellus 51B, 849–861.

Murphy, P.P., Harrison, D.E., Feely, R.A., Takahashi, T.T.,

Weiss, R.F., Gammon, R.H., 1998. Variability of DpCO2 in

the subarctic North Pacific: a comparison of results from

four expeditions. Tellus 50B, 185–204.

Nojiri, Y., 2001a. Surface water and atmospheric pCO2 data

observed aboard the Alligator Hope. Center for Global

Environment Research, National Institute for Environmen-

tal Studies, Japan, /www.mirc.jha.or.jp/nies/AHS.Nojiri, Y., 2001b. Surface water and atmospheric pCO2 data

observed aboard the Skaugran. Center for Global Environ-

ment Research, National Institute for Environmental

Studies, Japan, /www.mirc.jha.or.jp/nies/SKS.Parslow, J.S., 1981. Phytoplankton-zooplankton interactions:

Data analysis and modelling (with particular reference to

Ocean Station P (501N, 1451W) and controlled ecosystem

experiments). Ph.D. Thesis, University of British Columbia.

Quay, P., 1987. Was a carbon balance measured in the

equatorial pacific? Deep-Sea Research II 44, 1765–1782.

Rubin, S.I., 2000. Processes controlling the distribution of

carbon dioxide and nutrients in the South Pacific and

Southern Oceans. Ph.D. Thesis, Columbia University, New

York, NY, 223pp.

Rubin, S.I., Takahashi, T., Chipman, D.W., Goddard, J.G.,

1998. Primary production and nutrient utilization ratios in

the Pacific Sector of the Southern Ocean based on seasonal

changes in seawater chemistry. Deep-Sea Research I 45,

1211–1234.

Sabine, C.L., Key, R.M., 1998. Final report for underway

surface CO2 measurements on WOCE Leg P17E/P19S,

Technical Report #98-3, Ocean Tracer Laboratory, Depart-

ment of Geosciences, Princeton University, Princeton, NJ,

13pp.

Sabine, C.L., Key, R.M., Kozyr, A. (Eds.), 1997. Surface water

and atmospheric underway carbon dioxide data obtained

during the World Ocean Circulation Experiment Indian

Ocean Survey Cruises (R/V Knorr, December 1994–

January, 1996), ORNL/CDIAC-103, NDP-064, Carbon

Dioxide Information Analysis Center, Oak Ridge National

Laboratory, Oak Ridge, TN, 89pp.

Sabine, C.L., Key, R.M., Hall, M., Kozyr, A. (Eds.), 1999.

Carbon dioxide, hydrographic and chemical data obtained

during the R/V Thomas G. Thompson cruise in the

Pacific Ocean (WOCE section P10, October 5–November

10, 1993). ORNL/CDIAC-122, NDP-071, Carbon Dioxide

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–1622 1621

Information Analysis Center, Oak Ridge National Labora-

tory, US Department of Energy, Oak Ridge, TN, 68pp.

Sabine, C.L., Wanninkhof, R., Key, R.M., Goyet, C.,

Millero, F.J., 2000. Seasonal CO2 fluxes in the

tropical and subtropical Indian Ocean. Marine Chemistry

72, 33–55.

Sarmiento, J.L., Orr, J.C., Siegenthaler, U., 1992. A perturba-

tion simulation of CO2 uptake in an ocean general

circulation model. Journal of Geophysical Research 97,

3621–3645.

Siegenthaler, U., Sarmiento, J.L., 1993. Atmospheric carbon

dioxide and the ocean. Nature 365, 119–125.

Sweeney, C., 2000. Biogeochemical regimes and meso-scale

variability in the Ross Sea, Antarctica. Ph.D. Thesis,

Columbia University, New York, NY, 137pp.

Sweeney, C., Hansell, D.A., Carlson, C.A., Codispoti, L.A.,

Gordon, L.I., Marra, J., Millero, F.J., Smith, W.O.,

Takahashi, T., 2000b. Biogeochemical regimes, net commu-

nity production and carbon export in the Ross Sea,

Antarctica. Deep-Sea Research II 47, 3369–3394.

Sweeney, C., Smith, W.O., Hales, B., Bidigare, R.R., Carlson,

C.A., Codispoti, L.A., Gordon, L.I., Hansell, D., Millero,

F.J., Park, Mi-OK, Takahashi, T., 2000b. Nutrient and

carbon removal ratios and fluxes in the Ross Sea,

Antarctica. Deep-Sea Research II, 47, 3395–3421.

Takahashi, T., Goddard, J., 1998. Measurement of the total

CO2 concentration and total alkalinity in seawater collected

during the SONNE Cruise 102, May–June, 1985. Final

Technical Report, Lamont–Doherty Earth Observatory,

Palisades, NY, 54pp.

Takahashi, T., Olafsson, J., Goddard, J., Chipman, D.W.,

Sutherland, S.C., 1993. Seasonal variation of CO2 and

nutrients in the high-latitude surface oceans: a comparative

study. Global Biogeochemical Cycles 7, 843–878.

Takahashi, T., Takahashi, T.T., Sutherland, S.C., 1995. An

assessment of the role of the North Atlantic as a CO2 sink.

Philosophical Transactions of the Royal Society of London,

Series B 348, 143–152.

Takahashi, T., Feely, R.A., Weiss, R., Wanninkhof, R.H.,

Chipman, D.W., Sutherland, S.C., Takahashi, T.T., 1997.

Global air–sea flux of CO2: an estimate based on measure-

ments of sea–air pCO2 difference. Proceedings of the

National Academy of Science 94, 8292–8299.

Takahashi, T., Wanninkhof, R.H., Feely, R.A., Weiss, R.F.,

Chipman, D.W., Bates, N., Olafsson, J., Sabine, C.,

Sutherland, S.C., 1999. Net sea–air CO2 flux over the

global oceans: An improved estimate based on the sea–air

pCO2 difference. In: Yukihiro Nojiri (Ed.), Proceedings of

the Second International Symposium, CO2 in the Oceans

(ISSN 1341-4356), Center for Global Environmental

Research, National Institute for Environmental Studies,

Tsukuba, Japan, pp. 9–14.

Takahashi, T., Sweeney, C., Sutherland, S.C., Chipman, D.W.,

Goddard, J., Rubin, S.I., 2000. Method of underway pCO2

measurements in surface waters and the atmosphere during

the AESOP Expeditions, 1996–1998 in the Pacific sector of

the Southern Ocean and the Ross Sea. US JGOFS Data

Center, Woods Hole Oceanographic Institution, Woods

Hole, MA.

Thiele, G., Roether, W., Schlosser, P., Kuntz, R., Siedler, G.,

Stramma, L., 1986. Baroclinic flow and transient-tracer

fields in the Canary-Cape Verde Basin. Journal of Physical

Oceanography 16, 814–826.

Toggweiler, J.R., Dixon, K., Bryan, K., 1989. Simulations of

radiocarbon in a coarse resolution world ocean model I:

steady state pre-bomb distributions. Journal of Geophysical

Research 94, 8217–8242.

Wanninkhof, R., 1992. Relationship between wind speed

and gas exchange. Journal of Geophysical Research 97,

7373–7382.

Wanninkhof, R., McGillis, W.M., 1999. A cubic relationship

between gas transfer and wind speed. Geophysical Research

Letters 26, 1889–1892.

Wanninkhof, R., Doney, S.C., Takahashi, T., McGillis, W.R.

(2002). The effect of using time-averaged winds on regional

air–sea CO2 fluxes. In: Donelan M., Drennan W., Saltzman

E., Wanninkhof R. (Eds.), Gas Transfer at Water

Surfaces, AGU Geophys. Monograph 127, Washington,

DC. 351–356.

Weiss, R.F., 1974. Carbon dioxide in water and seawater: the

solubility of a non-ideal gas. Marine Chemistry 2, 203–215.

Wong, C.S., Chan, Y.-H., 1991. Temporal variations in the

partial pressure and flux of CO2 at ocean station P in the

subarctic northeast pacific ocean. Tellus 43B, 206–223.

World Ocean Database, 1998. Version 2, the National

Oceanographic Data Center, NOAA, Washington, DC.

T. Takahashi et al. / Deep-Sea Research II 49 (2002) 1601–16221622